Microsoft Unveils Muse AI Model That Can Generate Gameplay, But Won’t Make Games for Developers

Image source: Ninja Theory

Microsoft has unveiled a first-of-its-kind AI model, Muse, that can create environments based on visuals or player actions by recognizing the 3D world, the physics of the game, and how users interact with it.

Muse was trained on seven years of real gameplay from Ninja Theory’s hero shooter Bleeding Edge. Microsoft showed off some gameplay snippets (see video below) to demonstrate the technology’s capabilities.

Muse can currently generate “complex gameplay sections” at low (300×180) resolution and 10 FPS based on 10 frames (1 second) of actual gameplay and user input data throughout the session.

Microsoft insists that Muse is not intended to generate entire games or replace developers, but is designed to help interactive entertainment creators discover new possibilities and unleash their creativity.

Microsoft highlights areas where Muse is being used to improve classic games and port them to current devices, prototype and test ideas, and add new AI-based experiences to existing projects.

The company plans to share AI tools and experiments with Xbox players and developers at earlier stages of research. Microsoft also promised not to force internal studios to use AI in production.

Short interactive AI games coming soon to Copilot Labs